Diagnose before you prescribe
Sit one full, timed, honest practice test. Then go through every lost mark and label it: didn't know it, knew it but ran out of time, misread the question, careless slip. Count the four categories. That distribution is your study plan, and it is almost never what you assumed.
The whole method is in error analysis, and it is worth more than any book you can buy, because it stops you spending a month on the wrong problem.
If it's timing
- Practise sections against a clock, always. Untimed practice trains a skill the exam does not test.
- Learn to abandon. The hardest question is worth the same as the easiest. Two minutes spent on a question you won't get is two easy questions you didn't reach — this single habit is worth more points than most content revision.
- Do the easy ones first, deliberately. Order is your choice, not the paper's.
If it's content
Then it is a memory problem, and memory problems have a known solution. Card the specific gaps — the formulae, the grammar rules, the vocabulary — and put them on an FSRS schedule so they come back before you forget them rather than after the test.
If it's misreading or slips
This is a process problem and revision cannot touch it. It is fixed by a checklist you actually execute: underline what's being asked, write the units, sanity-check the magnitude. Boring, unsatisfying, and where a lot of people's easiest points are sitting.